@greatlhd/ailo-desktop 1.0.0 → 1.0.1

package/src/index.ts

@@ -2,24 +2,35 @@
 /**
  * Ailo Desktop — 超级端点
  *
- * 集成桌面能力 + 飞书 + 钉钉 + QQ + 邮件。
- * 各平台按需配置：有完整配置才启动对应 handler 并上报对应 blueprint。
+ * 集成桌面能力 + 飞书。
+ * 各平台按需配置：有完整配置才启动对应 handler 并上报对应工具。
  * 未配置的平台不会被 Ailo 感知，不影响其他能力正常工作。
  *
- * 子命令：ailo-desktop init [--defaults] — 初始化 config.json 与 Skills 目录（见 cli.ts）。
+ * 运行模式：
+ * - 桌面模式（有显示器）：截图、浏览器、鼠标键盘 + 文件操作、代码执行、MCP
+ * - 服务器模式（无显示器）：文件操作、代码执行、MCP
+ *
+ * 子命令：ailo-desktop init [--defaults] [--config-dir <path>] — 初始化 config.json 与 Skills 目录（见 cli.ts）。
  */
 
 const [, , subcommand] = process.argv;
 if (subcommand === "init") {
   import("./cli.js")
-    .then(({ runInit }) =>
-      runInit(process.argv.includes("--defaults"))
+    .then(({ runInit }) => {
+      const args = process.argv.slice(2);
+      let useDefaults = false;
+      let configDir: string | undefined;
+      for (let i = 0; i < args.length; i++) {
+        if (args[i] === "--defaults") useDefaults = true;
+        else if ((args[i] === "--config-dir" || args[i] === "-c") && args[i + 1]) configDir = args[++i];
+      }
+      return runInit(useDefaults, configDir)
         .then(() => process.exit(0))
         .catch((e) => {
           console.error(e);
           process.exit(1);
-        }),
-    )
+        });
+    })
     .catch((e) => {
       console.error("init 加载失败:", e);
       process.exit(1);
@@ -27,7 +38,7 @@ if (subcommand === "init") {
 }
 
 import { runEndpoint, type EndpointContext } from "@greatlhd/ailo-endpoint-sdk";
-import type { ContentPart, ToolCapability } from "@greatlhd/ailo-endpoint-sdk";
+import type { ContentPart, ToolCapability, SkillMeta } from "@greatlhd/ailo-endpoint-sdk";
 import { inferMime, classifyMedia, mediaPart } from "@greatlhd/ailo-endpoint-sdk";
 import { createInterface } from "readline";
 import { dirname, join } from "path";
@@ -70,29 +81,35 @@ import {
 import { CONFIG_FILENAME, NO_SUBJECT } from "./constants.js";
 import { errMsg } from "./utils.js";
 
-const BLUEPRINTS_DIR = join(__dirname, "..", "..", "..", "blueprints");
-
-function parseArgs(): { port?: number; blueprintUrl?: string } {
+function parseArgs(): { port?: number; configDir?: string; server?: boolean } {
   const args = process.argv.slice(2);
   let port: number | undefined;
-  let blueprintUrl: string | undefined;
+  let configDir: string | undefined;
+  let server: boolean | undefined;
   for (let i = 0; i < args.length; i++) {
     if (args[i] === "--port" && args[i + 1]) {
       const p = Number(args[++i]);
       if (!Number.isNaN(p) && p > 0) port = p;
-    } else if (args[i] === "--blueprint-url" && args[i + 1]) blueprintUrl = args[++i];
+    } else if ((args[i] === "--config-dir" || args[i] === "-c") && args[i + 1]) configDir = args[++i];
+    else if (args[i] === "--server" || args[i] === "-s") server = true;
   }
-  return { port, blueprintUrl };
+  return { port, configDir, server };
 }
 
-const { port: CLI_PORT, blueprintUrl: CLI_BLUEPRINT_URL } = parseArgs();
+const { port: CLI_PORT, configDir: CLI_CONFIG_DIR, server: CLI_SERVER } = parseArgs();
+
+function hasDisplay(): boolean {
+  if (CLI_SERVER) return false;
+  if (process.platform === "darwin" || process.platform === "win32") {
+    return true; // macOS/Windows 默认有显示器
+  }
+  return !!process.env.DISPLAY || !!process.env.WAYLAND_DISPLAY;
+}
 
-const BLUEPRINT_URL = CLI_BLUEPRINT_URL ?? join(BLUEPRINTS_DIR, "desktop-agent.blueprint.md");
-const BLUEPRINT_WEBCHAT = join(BLUEPRINTS_DIR, "webchat.blueprint.md");
-const BLUEPRINT_FEISHU = join(BLUEPRINTS_DIR, "feishu.blueprint.md");
+const isDesktopMode = hasDisplay();
 
 // ──────────────────────────────────────────────────────────────
-// 配置加载：各平台 + email
+// 配置加载：各平台
 // ──────────────────────────────────────────────────────────────
 
 export interface FeishuConfig { appId: string; appSecret: string }
@@ -125,58 +142,70 @@ const mcpManager = new LocalMCPManager();
 let endpointCtx: EndpointContext | null = null;
 let webchatApi: { recordAiloReply: (text: string, participantName: string) => boolean } | null = null;
 let feishuHandler: FeishuHandler | null = null;
-let lastMcpToolSnapshot: Map<string, ToolCapability> = new Map();
 const desktopStateStore = new DesktopStateStore();
 
-function computeToolDiff(oldTools: Map<string, ToolCapability>, newTools: ToolCapability[]): {
-  register: ToolCapability[];
-  unregister: string[];
-} {
-  const newMap = new Map(newTools.map((t) => [t.name, t]));
-  const register: ToolCapability[] = [];
-  const unregister: string[] = [];
-  for (const [name] of oldTools) {
-    if (!newMap.has(name)) unregister.push(name);
-  }
-  for (const [name, tool] of newMap) {
-    if (!oldTools.has(name)) register.push(tool);
-  }
-  return { register, unregister };
-}
+// 已上报的工具和技能（供 config_server API 使用）
+const reportedEndpointTools: ToolCapability[] = [];
+const reportedEndpointSkills: SkillMeta[] = [];
 
-async function syncMcpToolsToServer(): Promise<void> {
-  if (!endpointCtx) return;
-  const currentTools = mcpManager.getAllPrivateTools();
-  const { register, unregister } = computeToolDiff(lastMcpToolSnapshot, currentTools);
-  if (register.length === 0 && unregister.length === 0) return;
-  try {
-    await endpointCtx.update({
-      register: register.length > 0 ? { tools: register } : undefined,
-      unregister: unregister.length > 0 ? { tools: unregister } : undefined,
-    });
-    lastMcpToolSnapshot = new Map(currentTools.map((t) => [t.name, t]));
-    console.log(`[desktop] MCP 工具增量同步: +${register.length} -${unregister.length}`);
-  } catch (e: unknown) {
-    console.error("[desktop] MCP 工具增量同步失败:", errMsg(e));
+// ──────────────────────────────────────────────────────────────
+// 端点工具定义（上报给 Ailo）
+// ──────────────────────────────────────────────────────────────
+
+function buildEndpointTools(): ToolCapability[] {
+  const tools: ToolCapability[] = [];
+
+  // 文件操作工具
+  const fsToolDefs: [string, string][] = [
+    ["read_file", "读取文件内容。传入文件路径，返回文件内容。"],
+    ["write_file", "创建或覆写文件。将内容写入指定路径的文件。"],
+    ["edit_file", "编辑文件。在指定位置插入内容。"],
+    ["append_file", "追加内容。将内容追加到文件末尾。"],
+    ["list_directory", "列出目录内容。返回指定目录下的文件和子目录列表。"],
+    ["find_files", "查找文件。根据模式匹配文件名。"],
+    ["search_content", "搜索内容。在文件中搜索匹配的文本行。"],
+    ["delete_file", "删除文件。删除指定路径的文件。"],
+    ["move_file", "移动文件。将文件从源路径移动到目标路径。"],
+    ["copy_file", "复制文件。将文件从源路径复制到目标路径。"],
+  ];
+  for (const [name, desc] of fsToolDefs) {
+    tools.push({ name, description: desc });
   }
-}
 
-async function initSubsystems(): Promise<void> {
-  try {
-    await mcpManager.init();
-    mcpManager.startWatching();
-    let mcpSyncTimer: ReturnType<typeof setTimeout> | null = null;
-    mcpManager.setOnToolsChanged(() => {
-      if (mcpSyncTimer) return; // 防抖：500ms 内只同步一次
-      mcpSyncTimer = setTimeout(() => {
-        mcpSyncTimer = null;
-        console.log("[desktop] MCP 工具变更，增量同步到 Ailo");
-        syncMcpToolsToServer();
-      }, 500);
-    });
-  } catch (e: unknown) {
-    console.error("[desktop] MCP 初始化失败:", errMsg(e));
+  // 代码执行工具
+  tools.push(
+    { name: "execute_code", description: "执行代码。支持多种编程语言的代码执行。" },
+    { name: "exec", description: "执行 shell 命令。在服务器上执行 shell 命令并返回输出。" },
+  );
+
+  // MCP 管理工具
+  tools.push({
+    name: "mcp_manage",
+    description: "管理 MCP 服务器。列出、启动、停止 MCP 服务器，以及查看服务器提供的工具。",
+  });
+
+  // 飞书发送工具（飞书配置后可用）
+  tools.push({
+    name: "feishu_send",
+    description: "通过飞书发送消息。支持发送给指定 chat_id，可附带文本和文件附件。",
+  });
+
+  // 网页聊天发送工具
+  tools.push({
+    name: "webchat_send",
+    description: "通过网页聊天界面发送消息给用户。需要在浏览器中打开 Ailo Desktop 配置页。",
+  });
+
+  // 桌面模式工具
+  if (isDesktopMode) {
+    tools.push(
+      { name: "screenshot", description: "截取屏幕截图。可指定窗口或全屏。" },
+      { name: "browser_use", description: "浏览器自动化。控制浏览器进行网页操作。" },
+      { name: "mouse_keyboard", description: "鼠标键盘控制。模拟鼠标点击、键盘输入等操作。" },
+    );
   }
+
+  return tools;
 }
 
 // ──────────────────────────────────────────────────────────────
@@ -201,13 +230,38 @@ async function stopPlatformHandlers(): Promise<void> {
 }
 
 // ──────────────────────────────────────────────────────────────
-// 动态计算 blueprints（只注册已配置的平台）
+// 同步 MCP 工具到服务端
 // ──────────────────────────────────────────────────────────────
 
-function buildBlueprints(configPath: string): string[] {
-  const list: string[] = [BLUEPRINT_URL, BLUEPRINT_WEBCHAT];
-  if (loadFeishuConfig(configPath)) list.push(BLUEPRINT_FEISHU);
-  return list;
+async function syncMcpToolsToServer(): Promise<void> {
+  if (!endpointCtx) return;
+  const mcpTools = mcpManager.getAllPrivateTools();
+  try {
+    await endpointCtx.update({
+      register: { mcpTools },
+    });
+    console.log(`[desktop] MCP 工具同步: ${mcpTools.length} 个工具`);
+  } catch (e: unknown) {
+    console.error("[desktop] MCP 工具同步失败:", errMsg(e));
+  }
+}
+
+async function initSubsystems(): Promise<void> {
+  try {
+    await mcpManager.init();
+    mcpManager.startWatching();
+    let mcpSyncTimer: ReturnType<typeof setTimeout> | null = null;
+    mcpManager.setOnToolsChanged(() => {
+      if (mcpSyncTimer) return;
+      mcpSyncTimer = setTimeout(() => {
+        mcpSyncTimer = null;
+        console.log("[desktop] MCP 工具变更，同步到 Ailo");
+        syncMcpToolsToServer();
+      }, 500);
+    });
+  } catch (e: unknown) {
+    console.error("[desktop] MCP 初始化失败:", errMsg(e));
+  }
 }
 
 // ──────────────────────────────────────────────────────────────
@@ -227,71 +281,78 @@ function requireHandler<T>(handler: T | null, platform: string): T {
 const requireFeishu = () => requireHandler(feishuHandler, "飞书");
 
 function buildToolHandlers(): Record<string, (args: Record<string, unknown>) => Promise<ContentPart[] | unknown>> {
-  const handlers: Record<string, (args: Record<string, unknown>) => Promise<ContentPart[] | unknown>> = {
-    screenshot: async (args) => {
+  const handlers: Record<string, (args: Record<string, unknown>) => Promise<ContentPart[] | unknown>> = {};
+
+  // 桌面模式工具（需要显示器）
+  if (isDesktopMode) {
+    handlers.screenshot = async (args) => {
       const result = await captureDesktopObservation({
         capture_window: !!args.capture_window,
         screen: args.screen as number | "all" | undefined,
       });
       if (result.observation) desktopStateStore.saveObservation(result.observation);
       return result.parts;
-    },
-    browser_use: async (args) => browserUse(args),
-    execute_code: async (args) => {
-      if (!endpointCtx) throw new Error("端点未就绪");
-      return executeCode(endpointCtx, args);
-    },
-    exec: async (args) => {
-      if (!endpointCtx) throw new Error("端点未就绪");
-      return execTool(endpointCtx, args);
-    },
-    mouse_keyboard: async (args) => mouseKeyboard(args, { stateStore: desktopStateStore }),
-    mcp_manage: async (args) => {
-      const result = await mcpManager.handle(args);
-      if (result.toolsChanged) syncMcpToolsToServer();
-      return result.text;
-    },
-    // 网页聊天发送
-    webchat_send: async (args) => {
-      const text = args.text as string | undefined;
-      const participantName = args.participantName as string;
-      const attachments = (args.attachments as { path?: string }[]) ?? [];
-      if (!webchatApi && attachments.length === 0) {
-        return [{ type: "text", text: JSON.stringify({ ok: false, error: "webchat未就绪，请先连接Ailo并打开配置页" }) }];
-      }
-      const results: string[] = [];
-      // 发送文字
-      if (text && webchatApi) {
-        const ok = webchatApi.recordAiloReply(text, participantName ?? "");
-        results.push(ok ? "文字已发送" : "文字发送失败，请确认participantName与网页聊天中的称呼一致且用户在线");
-      }
-      // 发送附件
-      if (attachments.length > 0) {
-        if (!endpointCtx) return [{ type: "text", text: JSON.stringify({ ok: false, error: "端点未就绪，无法发送附件" }) }];
-        for (const att of attachments) {
-          if (att.path) {
-            await endpointCtx.sendFile(att.path);
-            results.push(`文件已发送：${att.path}`);
-          }
+    };
+    handlers.browser_use = async (args) => browserUse(args);
+    handlers.mouse_keyboard = async (args) => mouseKeyboard(args, { stateStore: desktopStateStore });
+  }
+
+  // 通用工具（桌面和服务器都可用）
+  handlers.execute_code = async (args) => {
+    if (!endpointCtx) throw new Error("端点未就绪");
+    return executeCode(endpointCtx, args);
+  };
+  handlers.exec = async (args) => {
+    if (!endpointCtx) throw new Error("端点未就绪");
+    return execTool(endpointCtx, args);
+  };
+  handlers.mcp_manage = async (args) => {
+    const result = await mcpManager.handle(args);
+    if (result.toolsChanged) syncMcpToolsToServer();
+    return result.text;
+  };
+
+  // 网页聊天发送
+  handlers.webchat_send = async (args) => {
+    const text = args.text as string | undefined;
+    const participantName = args.participantName as string;
+    const attachments = (args.attachments as { path?: string }[]) ?? [];
+    if (!webchatApi && attachments.length === 0) {
+      return [{ type: "text", text: JSON.stringify({ ok: false, error: "webchat未就绪，请先连接Ailo并打开配置页" }) }];
+    }
+    const results: string[] = [];
+    if (text && webchatApi) {
+      const ok = webchatApi.recordAiloReply(text, participantName ?? "");
+      results.push(ok ? "文字已发送" : "文字发送失败，请确认participantName与网页聊天中的称呼一致且用户在线");
+    }
+    if (attachments.length > 0) {
+      if (!endpointCtx) return [{ type: "text", text: JSON.stringify({ ok: false, error: "端点未就绪，无法发送附件" }) }];
+      for (const att of attachments) {
+        if (att.path) {
+          await endpointCtx.sendFile(att.path);
+          results.push(`文件已发送：${att.path}`);
         }
       }
-      if (results.length === 0) return [{ type: "text", text: JSON.stringify({ ok: false, error: "请提供text或attachments" }) }];
-      return [{ type: "text", text: JSON.stringify({ ok: true, results }) }];
-    },
-    // 飞书发送
-    feishu_send: async (args) => {
-      const h = requireFeishu();
-      const atts = ((args.attachments as any[]) ?? []).map((a: any) => ({
-        type: a.type,
-        file_path: a.path,
-        mime: a.mime,
-        name: a.name,
-        duration: a.duration,
-      }));
-      await h.sendText(args.chat_id as string, (args.text as string) ?? "", atts);
-      return [{ type: "text", text: `已发送到 ${args.chat_id}` }];
-    },
+    }
+    if (results.length === 0) return [{ type: "text", text: JSON.stringify({ ok: false, error: "请提供text或attachments" }) }];
+    return [{ type: "text", text: JSON.stringify({ ok: true, results }) }];
+  };
+
+  // 飞书发送
+  handlers.feishu_send = async (args) => {
+    const h = requireFeishu();
+    const atts = ((args.attachments as any[]) ?? []).map((a: any) => ({
+      type: a.type,
+      file_path: a.path,
+      mime: a.mime,
+      name: a.name,
+      duration: a.duration,
+    }));
+    await h.sendText(args.chat_id as string, (args.text as string) ?? "", atts);
+    return [{ type: "text", text: `已发送到 ${args.chat_id}` }];
   };
+
+  // 文件系统工具
   for (const name of FS_TOOLS) {
     handlers[name] = async (args) => fsTool(name, args);
   }
@@ -304,38 +365,19 @@ function buildToolHandlers(): Record<string, (args: Record<string, unknown>) =>
 
 async function main(): Promise<void> {
   const port = CLI_PORT ?? (await promptPort());
-  const configPath = join(process.cwd(), CONFIG_FILENAME);
+  const configDir = CLI_CONFIG_DIR ?? process.cwd();
+  const configPath = join(configDir, CONFIG_FILENAME);
+  console.log(`[desktop] 配置目录: ${configDir}`);
+  console.log(`[desktop] 运行模式: ${isDesktopMode ? "桌面模式" : "服务器模式"}`);
   const connectionState = { connected: false, endpointId: "" };
   let webchatCtxRef: EndpointContext | null = null;
   let connectAttempt = 0;
   let endpointConnecting = false;
 
-  async function reloadPlatformHandler<
-    C,
-    H extends { start(ctx: EndpointContext): Promise<void>; stop(): Promise<void> },
-  >(opts: {
-    getHandler: () => H | null;
-    setHandler: (h: H | null) => void;
-    loadConfig: () => C | null;
-    createHandler: (cfg: C) => H;
-    blueprint: string;
-  }): Promise<void> {
-    if (!endpointCtx) return;
-    const wasRunning = !!opts.getHandler();
-    if (opts.getHandler()) {
-      await opts.getHandler()!.stop();
-      opts.setHandler(null);
-    }
-    const cfg = opts.loadConfig();
-    if (cfg) {
-      const h = opts.createHandler(cfg);
-      opts.setHandler(h);
-      await h.start(endpointCtx);
-      if (!wasRunning) await endpointCtx.update({ register: { blueprints: [opts.blueprint] } });
-    } else if (wasRunning) {
-      await endpointCtx.update({ unregister: { blueprints: [opts.blueprint] } });
-    }
-  }
+  // 端点描述
+  const endpointInstructions = isDesktopMode
+    ? "Ailo Desktop 端点 - 具备文件操作、代码执行、截图、浏览器自动化、鼠标键盘控制等能力"
+    : "Ailo Server 端点 - 具备文件操作、代码执行等服务器能力";
 
   async function applyConnectionConfig(overrides?: AiloConnectionConfig): Promise<void> {
     const cfg = overrides ?? loadConnectionConfig(configPath);
@@ -344,8 +386,13 @@ async function main(): Promise<void> {
 
     endpointConnecting = true;
     connectAttempt = 0;
+
+    // 填充已上报的工具列表
+    const builtInTools = buildEndpointTools();
+    reportedEndpointTools.length = 0;
+    reportedEndpointTools.push(...builtInTools);
+
     try {
-      const blueprints = buildBlueprints(configPath);
       runEndpoint({
         ailoWsUrl: cfg.url,
         ailoApiKey: cfg.apiKey,
@@ -358,10 +405,11 @@ async function main(): Promise<void> {
             connectionState.connected = true;
             connectionState.endpointId = cfg.endpointId;
             webchatCtxRef = ctx;
-            lastMcpToolSnapshot = new Map(mcpManager.getAllPrivateTools().map((t) => [t.name, t]));
             serverRef.notifyContextAttached();
             await startPlatformHandlers(ctx, configPath);
-            console.log("[desktop] 桌面端点已启动");
+            // 同步 MCP 工具
+            await syncMcpToolsToServer();
+            console.log("[desktop] 端点已启动");
           },
           stop: async () => {
             await stopPlatformHandlers();
@@ -370,14 +418,17 @@ async function main(): Promise<void> {
             connectionState.connected = false;
             connectionState.endpointId = "";
             webchatCtxRef = null;
+            reportedEndpointTools.length = 0;
+            reportedEndpointSkills.length = 0;
             await mcpManager.shutdown();
-            await stopBrowser();
-            console.log("[desktop] 桌面端点已停止");
+            if (isDesktopMode) await stopBrowser();
+            console.log("[desktop] 端点已停止");
           },
         },
         caps: ["message", "tool_execute"],
-        blueprints,
-        tools: mcpManager.getAllPrivateTools(),
+        tools: builtInTools,
+        mcpTools: mcpManager.getAllPrivateTools(),
+        instructions: endpointInstructions,
         toolHandlers: buildToolHandlers(),
         onUnknownTool: async (name: string, args: Record<string, unknown>) => {
           const idx = name.indexOf(":");
@@ -400,11 +451,7 @@ async function main(): Promise<void> {
             return;
           }
           try {
-            await client.reconnect(undefined, {
-              url: latest.url,
-              apiKey: latest.apiKey,
-              endpointId: latest.endpointId,
-            });
+            await client.reconnect(undefined, undefined, undefined, mcpManager.getAllPrivateTools());
           } catch (e) {
             console.error("[desktop] 重试连接失败:", errMsg(e));
           }
@@ -422,15 +469,15 @@ async function main(): Promise<void> {
     mcpManager,
     getConnectionStatus: () => connectionState,
     getWebchatCtx: () => webchatCtxRef,
+    getEndpointCtx: () => endpointCtx,
+    getEndpointTools: () => reportedEndpointTools.map(t => ({ name: t.name, description: t.description ?? "" })),
+    getEndpointSkills: () => reportedEndpointSkills.map(s => ({ name: s.name, description: s.description ?? "" })),
     port,
     configPath,
-    blueprintUrl: BLUEPRINT_URL,
-    blueprintLocalPath: join(BLUEPRINTS_DIR, "desktop-agent.blueprint.md"),
-    getBlueprintPaths: () => buildBlueprints(configPath),
     onWebchatReady: (api) => { webchatApi = api; },
     onRequestReconnect: async () => {
       if (!endpointCtx) return;
-      await endpointCtx.client.reconnect([]);
+      await endpointCtx.client.reconnect(undefined, undefined, undefined, mcpManager.getAllPrivateTools());
     },
     onConnectionConfigSaved: async (config) => {
       const cfg: AiloConnectionConfig = {
@@ -439,18 +486,15 @@ async function main(): Promise<void> {
         endpointId: config.endpointId,
       };
       if (endpointCtx) {
-        await endpointCtx.client.reconnect(undefined, cfg);
+        await endpointCtx.client.reconnect(undefined, cfg, undefined, mcpManager.getAllPrivateTools());
       } else {
         await applyConnectionConfig(cfg);
       }
     },
-    onFeishuConfigSaved: () => reloadPlatformHandler({
-      getHandler: () => feishuHandler,
-      setHandler: (h) => { feishuHandler = h; },
-      loadConfig: () => loadFeishuConfig(configPath),
-      createHandler: (cfg) => new FeishuHandler(cfg),
-      blueprint: BLUEPRINT_FEISHU,
-    }),
+    onFeishuConfigSaved: async () => {
+      // 飞书配置变更时，重新同步 MCP 工具（可能包含新的飞书工具）
+      await syncMcpToolsToServer();
+    },
     getFeishuStatus: () => ({
       configured: !!loadFeishuConfig(configPath),
       running: !!feishuHandler,